Information Processing Apparatus, Information Processing Method, and Program

Abstract

To provide technology which is easier to use for even a user unfamiliar with statistics or the like, and can provide information useful for a manufacturing business or the like.SOLUTION: A learning data management part 60 acquires raw data transmitted from a user terminal 2. A process chart generation part 62 presents an image for dividing a manufacturing step of an article into a plurality of steps. A process chart presentation part 82 generates various images about the process chart on the basis of contents of learning data stored in a learning DB 300, and presents the images to a user. A data item reception part 90 receives selections of data items about each of the plurality of steps from the user who presents the various images about the process chart. A learning processing part 68 executes processing about statistical machine learning to the learning data stored in the learning data DB 300, and generates a prediction model reflecting statistical quality in the learning data, various algorithms, or the like as learning results.SELECTED DRAWING: Figure 3

Description

Technical Field

[0001] The present invention relates to an information processing apparatus, an information processing method, and a program.

Background Art

[0002] Conventionally, in the manufacturing industry and the like, it has not been easy to determine the characteristics of products suitable for the purpose, the conditions of the optimal manufacturing process of the products, and the like. Therefore, in order to determine the optimal manufacturing conditions, a method that requires a great deal of cost and time, such as creating and prototyping many prototypes, has been adopted. For example, Patent Document 1 proposes a method of determining manufacturing conditions suitable for the manufacture of metal products by calculating and outputting the characteristics of intermediate products at stages after passing through each of a plurality of manufacturing processes for manufacturing metal products using a group of mathematical formulas.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, according to the conventional technology including the technology described in Patent Document 1 above, it is only possible to calculate a predicted value indicating the characteristics of a metal product based on the values output for each of a plurality of manufacturing processes, and it is difficult to handle for those who are not good at statistical processing using mathematical formulas and the like. Furthermore, it is also insufficient for determining manufacturing conditions corresponding to various situations in the manufacturing industry.

[0005] Also, as a method for solving such problems, a method using machine learning or the like can be considered. However, even when using machine learning or the like, the specifications will inevitably be very difficult to handle for users who are not accustomed to statistical processing, such as engineers who have been engaged in conventional manufacturing, in terms of setting and selecting various parameters. For details of the method for assisting in determining manufacturing conditions using machine learning or the like, refer to Japanese Patent Application Nos. 2022-74143 and 2022-74144 filed by the present applicant.

[0006] The present invention has been made in view of such a situation, and an object thereof is to provide a technology that is easier to use for users who are not accustomed to statistics or the like and can provide useful information for manufacturing industries and the like.

[0007] To achieve the above object, an information processing apparatus according to an aspect of the present invention is an information processing apparatus that can be used for statistical learning processing related to the manufacture of an article, a process table generation unit (for example, process table generation unit 62) that generates a process table, which is an image related to the manufacturing process of the article, based on an input from a user; a data item setting unit (for example, data item reception unit 90) that sets data items related to the learning for each of the processes included in the process table; a learning condition determination unit (for example, learning condition determination unit 64) that determines conditions for the learning process based on the contents of the set data items; a learning processing unit (for example, learning processing unit 68) that performs the learning process based on the learning conditions determined by the learning condition determination unit and outputs the result of the learning process; and includes.

[0008] An information processing method and program according to an aspect of the present invention are also provided as an information processing method or program corresponding to the information processing apparatus according to an aspect of the present invention.

Effects of the Invention

[0009] According to the present invention, it is possible to provide a technology that is more user-friendly and can provide useful information for the manufacturing industry and the like, even for users who are not familiar with statistics and the like.

Brief Description of the Drawings

[0010]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Embodiments for Carrying Out the Invention

[0011] <Description of the Summary> Hereinafter, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a diagram showing the configuration of an information processing system according to an embodiment of the present invention.

[0012] Before explaining FIG. 1, the background for constructing an information processing system according to an embodiment of the present invention will be briefly described. As described above, the applicant of the present invention has previously proposed a method for assisting in determining the manufacturing conditions of an article using statistical methods such as machine learning (see Japanese Patent Application No. 2022-74143, Japanese Patent Application No. 2022-74144, etc.). However, when performing data analysis in a field that requires specialized knowledge, such as the manufacturing industry, both knowledge of data analysis and domain knowledge of the specialized field are required. Therefore, rather than simply generating a machine learning engine as a finished product, it is desirable for on-site engineers to be able to perform data analysis themselves and make appropriate trial-and-error improvements when the accuracy is poor. The present invention has been made in view of such circumstances, and proposes a method for assisting in determining the manufacturing conditions of an article using statistical methods such as machine learning that can be more easily handled.

[0013] As shown in FIG. 1, this system includes a server 1 and a user device 2. The server 1 and the user device 2 are interconnected via a predetermined network N such as the Internet. Note that the network N is not an essential component, and for example, NFC (Near Field Communication), Bluetooth (registered trademark), LAN (Local Area Network), etc. may be used.

[0014] <Hardware Configuration> FIG. 2 is a block diagram showing an example of the hardware configuration of the server in the information processing system of FIG. 1. The server 1 is configured by a personal computer or the like. As shown in FIG. 2, the server 1 includes a control unit 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, a bus 14, an input / output interface 15, an output unit 16, an input unit 17, a storage unit 18, a communication unit 19, and a drive 20.

[0015] The control unit 11 is composed of a microcomputer including a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), and a semiconductor memory, and executes various processes according to a program recorded in the ROM 12 or a program loaded from the storage unit 18 to the RAM 13. The RAM 13 also appropriately stores information and the like necessary for the control unit 11 to execute various processes.

[0016] The control unit 11, the ROM 12, and the RAM 13 are interconnected via a bus 14. An input / output interface 15 is also connected to this bus 14. The output unit 16, the input unit 17, the storage unit 18, the communication unit 19, and the drive 20 are connected to the input / output interface 15.

[0017] The output unit 16 is composed of various liquid crystal displays, speakers, etc., and outputs various information as images and sounds.

[0018] The input unit 17 is composed of a keyboard, a mouse, etc., and inputs various information.

[0019] The storage unit 18 is composed of an HDD (Hard Disk Drive), an SSD (Solid State Drive), etc., and stores various data. In this embodiment, for example, various information including various programs and various databases is stored.

[0020] The communication unit 19 controls communication with other devices (for example, the user device 2) via a network N including the Internet.

[0021] The drive 20 is provided as necessary. A removable medium 31 made of a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is appropriately mounted on the drive 20. The program read from the removable medium 31 by the drive 20 is installed in the storage unit 18 as necessary. The removable medium 31 can also store various data stored in the storage unit 18 in the same manner as the storage unit 18.

[0022] Note that since the hardware configuration of the user device 2 can be basically the same as that of the server 1, the description thereof is omitted here.

[0023] FIG. 3 is a block diagram showing an example of the functional configuration related to the parameter setting process among the functional configurations of the server in FIG. 2 and the user device in FIG. 1.

[0024] As shown in FIG. 3, the control unit 11 of the server 1 is provided with a learning data management unit 60, a process table generation unit 62, a learning condition determination unit 64, a correlation coefficient processing unit 66, and a learning processing unit 68. In addition, a learning data DB 300 and a learning result DB 400 are provided in an area of the storage unit 18 of the server 1. The learning data DB 300 stores various learning data generated by performing pre-processing for learning on raw data (numerical data measured for each data item, etc.) previously acquired from the user. The learning result DB 400 stores learning results (prediction models generated as a result of learning, various parameters, etc.) generated as a result of learning described later.

[0025] The learning data management unit 60 of the server 1 acquires the raw data transmitted from the user device 2. The learning data management unit 60 generates learning data based on the acquired raw data and stores the information in the learning data DB 300.

[0026] The process sheet generation unit 62 presents an image for classifying the manufacturing process of the article into a plurality of processes. Specifically, the process sheet generation unit 62 generates an image related to the manufacturing process for classifying and displaying the manufacturing process of a predetermined article into a plurality of manufacturing processes, and presents the image to the user via the user device 2. Thereby, the user can freely create a process sheet based on his or her own knowledge and experience. Note that the user can also create a new arbitrary process sheet and freely modify the created process sheet.

[0027] Here, the process sheet generation unit 62 is provided with a data item selection reception unit 80 and a process sheet presentation unit 82. The data item selection reception unit 80 receives the selection of "final inspection target value" and "final inspection result value" by the user. Here, the final inspection target value means the item value (for example, the quality of the product, etc.) that the user finally aims for when considering the manufacturing conditions of the article. On the other hand, the final inspection result value means the item value actually measured for the manufactured article (for example, the completed chocolate). Note that the definitions and names of the final inspection target value and the final inspection result value are merely examples and are not limited.

[0028] The process sheet presentation unit 82 generates various images related to the process sheet based on the content of the learning data stored in the learning data DB 300, and presents the images to the user. In addition, the process sheet presentation unit 82 receives information regarding an input operation from the user who has confirmed the various images related to the presented process sheet, and appropriately changes the content of the process sheet (adds or changes processes) according to the content. Note that the process sheet is a diagram or table for clearly displaying the manufacturing process for manufacturing a predetermined article, and the details will be described later with reference to FIGS. 5 and 6 and the like.

[0029] The learning condition determination unit 64 determines the conditions for learning described later based on the content of the data items received from the user. Specifically, the learning condition determination unit 64 is provided with a data item reception unit 90 and a non-use data item reception unit 92. The data item reception unit 90 receives the selection of data items for each of a plurality of processes from a user who has been presented with various images related to the process table. The non-use data item reception unit 92 receives the selection of data items not used in learning. Here, the data items mentioned refer to each of the predetermined measurement items preset in the learning data. In the learning data, the actual measurement results corresponding to each data item are stored in various databases and the like in the form of lists or tables. However, whether the actual measurement results are stored or not is arbitrary by the user or the like, and there may be data items for which the measurement results are not stored. The method for determining the learning conditions by the learning condition determination unit 64 will be described later with reference to FIGS. 6 to 8 and the like.

[0030] The correlation coefficient processing unit 66 calculates the correlation coefficient between predetermined data items (for example, the data to be predicted and the manufacturing conditions), and presents the information to the user via the user device 2. Note that the data items for which the correlation coefficient processing unit 66 calculates the correlation coefficient are not limited to the data items to be predicted and the manufacturing conditions, and the correlation coefficient between any other arbitrary data items may be calculated and presented.

[0031] Then, the learning condition determination unit 64 determines the final learning conditions based on the data items for which selection has been received by the data item reception unit 90, the data items for which selection has been received by the non-use data item reception unit 92, the input content of the user with respect to the calculation result of the correlation coefficient processing unit 66, and the like.

[0032] The learning processing unit 68 executes a learning process based on the learning data stored in the learning data DB 300 and the learning conditions determined by the learning condition determination unit 64. Specifically, the learning processing unit 68 executes processing related to statistical machine learning on the learning data stored in the learning data DB 300, and generates a prediction model and various algorithms that reflect the statistical properties in the learning data as learning results. Note that the method of statistical machine learning executed by the learning processing unit 68 is arbitrary. In this embodiment, for example, a learning result is generated by a method that combines a neural network, a regression model by sparse modeling, or the like.

[0033] Here, as shown in FIG. 3, a detailed functional configuration description of the user device 2 is omitted. Specifically, for example, the user device 2 executes processes such as displaying various images and the like transmitted from the server 1 on a display unit (not shown), receiving an input operation from the user, and transmitting the information to the server 1.

[0034] FIG. 4 is a diagram showing an example of an image displayed on the user device of FIG. 3. Specifically, FIG. 4 shows an example of an image displayed on the user device 2 regarding the selection of the final inspection target value and the final inspection result value.

[0035] In the example of FIG. 4, various parameters related to the chocolate manufacturing process are displayed as an example. Specifically, in the example of FIG. 4, as the final inspection target value, the data item of "target smoothness" is selected, and as the final inspection result value, the data item of "measured smoothness" is selected. Also, candidates for other data items that are not selected are displayed at the bottom of FIG. 4. The user can select any data item among the data items thus displayed and determine the data items of the final inspection target value and the final inspection result value by pressing the selection confirmation button K1. Note that each data item displayed here may be an arbitrary data item preset by this system, or may be a data item set in the raw data or learning data acquired from the user.

[0036] On the other hand, FIG. 5 shows an example of an image displayed on the user device 2 when assigning data items. The example in Fig. 5 shows a process table created according to the manufacturing process of chocolate, along with candidates for parameters that can be set for each process. Note that, as described above, the user can freely add or delete any process as needed while checking the image that serves as the basis for creating the process table or the image shown in Fig. 5. Specifically, in the example of Fig. 5, as an example of the chocolate manufacturing process, the following processes are set and displayed as a process table: material information at the start of the process (start), roasting cocoa (first process), crushing cocoa (second process), mixing (third process), atomization (fourth process), refining (fifth process), tempering (sixth process), and molding (final process). Furthermore, the user can set data items related to each process while checking the image shown in Fig. 5. Specifically, in the example of Fig. 5, the user can set the processing value (under what conditions to manufacture the product) and the intermediate inspection result value (the value actually measured as a result of manufacturing, the manufacturing conditions that cannot be controlled) related to each process.

[0037] Here, in this system, the data items set as processing values and the data items set as intermediate inspection result values are treated as clearly different concepts and may be used for learning. For example, this system can examine whether sufficient accuracy learning can be performed using only the data items set as processing values in learning, and perform learning using the data items set as intermediate inspection result values only when necessary. In the first place, when the system learns the relationship between the final inspection result value of an article and the manufacturing conditions, it is preferable to learn using only the manufacturing conditions that are as controllable as possible, rather than simply learning all the manufacturing conditions in parallel. In this regard, for example, the intermediate inspection result value in FIG. 5 is an observed value observed as a result of a previous process in the manufacturing process, and it is difficult to change it to an arbitrary value. Therefore, the system, for example, sets the intermediate inspection result value as the objective variable, sets the data items of each previous manufacturing process as explanatory variables, and learns the relationship between them, and if the accuracy of the learning is low, the corresponding intermediate inspection result value is kept, and if the accuracy of the learning is high, the intermediate inspection result value is excluded. By selecting the manufacturing conditions in this way, the system can provide a learning environment that is more user-friendly.

[0038] By performing these operations, the user can easily select the data items to be used in the study. In the example of Fig. 5, the user can select the data items related to each process and press the process chart creation button K2 to proceed to the next stage of the process chart creation. Note that the user does not need to select data items that are not used in the study, for example. The user can freely create any process schedule they want while checking the process schedule in the process of being created displayed on the screen in this way, and can also freely change the division and flow of each manufacturing process by adding or deleting manufacturing processes. Using the process chart created in this way, the user can easily allocate each data item to be used in learning freely to each process, delete data items that are not related to the manufacturing process, etc. Particularly in fields such as manufacturing, more accurate and efficient learning can be achieved by having a user who is familiar with the products and manufacturing processes of his / her company perform such processing.

[0039] The final process sheet generated in this way is displayed, for example, as an image shown in FIG. 6. Here, the user can confirm that there are no errors in the process sheet created by checking the image in FIG. 6 and can select any data item. In the example of FIG. 6, the data items of "Temperature during kneading" and "Kneading time (2)" in the sixth process are selected and are shown hatched. The data items selected in this way are set as the data items that the user wants to know on the selection screen of the data items used for learning shown in FIG. 7. Thereby, while checking whether the setting of each data item is correct while checking the content of the completed process sheet, the user can select the data items that the user wants to know and can determine the content of the process sheet by pressing the process sheet confirmation button K3. Note that the data items that the user wants to know here are the data items for which the user wants to know the optimal manufacturing conditions in achieving the target specifications (target smoothness). Simply put, the user can manufacture chocolate with the target specifications (target smoothness) by adjusting the values of the data items of "Temperature during kneading" and "Kneading time (2)" to optimal values.

[0040] FIG. 7 is a diagram showing an example of an image displayed on the user device in FIG. 3, and is a diagram showing an example different from the examples in FIGS. 4 to 6. Specifically, FIG. 7 is an example of an image displayed on the user device 2 when selecting the data items used for learning.

[0041] In the example of FIG. 7, as the data items required for learning, the data items of "Target smoothness", "Variety of cocoa", "Country of origin of cocoa", "Weight of cocoa", "Roasting time", "Roasting temperature", "Crushing time", "Amount of sugar", "Amount of powdered milk", and "Amount of cocoa butter" are selected. The data items displayed in this item are selected as the data items used for learning. If there are data items that are not required for learning among the data items selected here, the user can switch the use / non-use of each data item, for example, by pressing the H1 bar or the like. Also, in the example of FIG. 7, as the data items of the target to be known, that is, the data items to be the target of inference, the temperature at the time of refining and the refining time (2) in the sixth step are set. This shows the data items selected in the screen situation shown in FIG. 6. Also, in the example of FIG. 7, as the unusable data items, the data items of "type of mold", "temperature of refrigerator", "refrigeration time", "measured smoothness", and "type of chocolate" are displayed. In the example of FIG. 7, usually, the data items set in the steps before the data items selected in FIG. 6 are set as the data required for learning, and the data items set in the steps after the data items selected in FIG. 6 are set as the unusable data items. While checking such an image, the user can easily determine the data items to be used for learning by selecting the data items to be used for learning and pressing the data confirmation button K4.

[0042] Here, in many cases, the processing values and the like can be arbitrarily changed by the user or the like, while the intermediate inspection result values and the final inspection result values are the results realized by the processing, so there is no degree of freedom and they are determined according to the natural law. Therefore, in many conventional techniques, generally only the variables corresponding to the intermediate inspection result values and the final inspection result values in this system are set as the target variables, and only the variables corresponding to the processing values and the like are set as the explanatory variables. However, when production is carried out, there is a target value for the final inspection result value, so actually there is no infinite degree of freedom in the processing value either. Especially for the processing value in the final step, since it must converge to the value aiming at the final inspection result value with the processing value of that step, there is almost no degree of freedom similar to the intermediate inspection result value and the final inspection result value. Therefore, it can be said that the processing value in production has the same characteristics (the characteristic that there is no degree of freedom) as the inspection result value, so it is also possible to perform learning by setting the processing value as the target variable and setting the previous processing value, intermediate inspection result value, and final inspection target value as the explanatory variables. The inventors of the present invention have clarified and can utilize such properties. For the above reasons, this system can set any of the data items set as the above-mentioned processed values and the data items set as each inspection result value (intermediate inspection result value or final inspection result value) as the target (objective variable) of prediction based on the learning results. Specifically, for example, even the processed value in the intermediate process in the manufacturing process (the temperature during kneading in the 6th process in Fig. 5), etc., prediction can be made using the learning results. By performing such prediction, the user can, for example, measure some intermediate inspection result value in the process immediately before the final process, and flexibly utilize this system to adjust the manufacturing conditions, such as adjusting the processed value in the final process to an optimal value by himself based on that value and the processed values and intermediate inspection result values before that process.

[0043] FIG. 8 is a diagram showing an example of an image displayed on the user device in FIG. 3, and shows an example different from the examples in FIGS. 4 to 7. Specifically, in the example of FIG. 8, an example of an image displayed on the user device 2 is shown to confirm the detailed settings regarding learning.

[0044] In the example of FIG. 8, the allowable errors of "temperature during kneading" and "kneading time (2)", which are the targets of inference, and the correlation coefficients between "temperature during kneading", "kneading time (2)" and each data item are respectively displayed. Here, the allowable error is a criterion for accuracy evaluation based on the difference between the predicted value and the measured value of the model. The allowable error in this system does not simply have a statistical meaning, but can also be affected by, for example, the user's prior knowledge and subjectivity. For example, when a certain variable is evaluated as important by a user proficient in the manufacturing industry, the allowable error is determined by reflecting such preconditions regardless of the statistical variation. Note that the value of the allowable error can be arbitrarily changed by the user. That is, the allowable error of this system means the range (width) regarding whether the error between the measured value and the predicted value of the manufacturing condition variables can be regarded as the same manufacturing condition in this service, taking into account the user's prior knowledge, subjectivity, etc. Note that this standard may change due to updates of various information, the passage of time, etc. in order to consider the user's prior knowledge, subjectivity, etc. The correlation coefficient is an index that measures the strength of the linear relationship between each data item. Specifically, the closer the correlation coefficient is to 1, the higher the positive correlation is evaluated, and the closer it is to -1, the more negative correlation is evaluated. For example, in statistical machine learning, if data items with extremely high correlation coefficients are used in learning, there is a risk of reducing the learning accuracy. Therefore, when the user sees an extremely high correlation coefficient, the user can change one of the data items to a data item not used in learning. The user can confirm such detailed learning conditions and generate a learning result by pressing the learning model creation button K5.

[0045] Figure 9 is a flowchart for explaining the flow of data item setting processing executed by the server in Figure 3.

[0046] In step S1, the data item selection reception unit 80 receives the selection of "final inspection target value" and "final inspection result value" by the user.

[0047] In step S2, based on the content of the learning data stored in the learning data DB 300, the process table presentation unit 82 generates various images related to the process table and presents the images to the user via the user device 2.

[0048] In step S3, the process table presentation unit 82 receives information regarding the input operation from the user who has confirmed the various images related to the presented process table, and appropriately changes the content of the process table (adds or changes processes) according to the content.

[0049] In step S4, the data item reception unit 90 receives, from the user presented with various images related to the process chart, the selection of data items for each of a plurality of processes.

[0050] In step S5, the unused data item reception unit 92 receives the selection of data items not used in learning.

[0051] In step S6, the correlation coefficient processing unit 66 calculates the correlation coefficient between predetermined data items (for example, the data item to be predicted and the manufacturing conditions), and presents the information to the user via the user device 2.

[0052] In step S7, the learning condition determination unit 64 determines the final learning conditions based on the data items selected by the data item reception unit 90, the data items selected by the unused data item reception unit 92, the input content of the user with respect to the calculation result of the correlation coefficient processing unit 66, and the like.

[0053] In step S8, the learning processing unit 68 executes processing related to statistical machine learning on the learning data stored in the learning data DB 300, and generates a prediction model and various algorithms that reflect the statistical properties in the learning data as learning results. Thus, the data item setting process ends.

[0054] Although one embodiment of the present invention has been described above, the present invention is not limited to the above-described embodiment, and modifications, improvements, etc. within the scope that can achieve the object of the present invention are included in the present invention.

[0055] Summarizing the above, it is considered that the user has at least the following advantages by using this system. (1) This system can provide a UI (user interface) that, for example, does not make the operation of difficult AI (machine learning) seem difficult. Therefore, engineers in the manufacturing industry and others can easily utilize the technology of AI (machine learning) to obtain the effects of data analysis. Specifically, by using this system, the following advantages are considered to exist for users. (a) The user can easily set, for example, the target variable and explanatory variables. For example, the user can also automatically set the target variable and explanatory variables. (b) The user can easily perform operations such as selecting necessary items from among the candidates for explanatory variables. (c) The user can easily perform learning while observing the minimum rules and conditions required for learning (for example, not setting data items in processes after the target variable as explanatory variables). For example, the user can automatically set up a learning environment that observes the rules and conditions required for learning just by operating according to the flow of this system. This can also be done. (2) This system can reflect the domain knowledge of engineers in the manufacturing industry and others (especially domain knowledge related to the manufacture of a predetermined article) in learning. Therefore, engineers in the manufacturing industry and others can obtain efficient and accurate learning results. That is, it is considered that the user can omit unnecessary learning processes and perform efficient and highly accurate learning, for example, by setting variables that can be inferred not to be involved as unused for the target variable and others based on the experience of manufacturing articles so far. (3) This system can reflect the domain knowledge of engineers in the manufacturing industry and others (especially domain knowledge that can be used across different industries in manufacturing) in learning. Therefore, a wide range of users can use it without being restricted to a specific industry or product. That is, manufacturing processes exist in many fields of manufacturing, and manufacturing personnel have certain common knowledge regarding manufacturing processes regardless of what products they are manufacturing. And in many manufacturing sites, for example, it is considered that "QC process sheets" (so-called process sheets) or similar items are being used. Therefore, users can use this system without being restricted to specific industries or products, and when using this system, it is considered to have a high affinity with the "QC process sheets" and the like that were conventionally used, and can be easily acquired and used. Note that the "QC process sheet" mentioned here is merely an example, and it may also be a diagram, table, document, etc. that shows the process of work. (4) By being organized as a process sheet, this system can obtain an overview perspective, and although it is not digitized for reasons such as not being measured in the current production, it can make it easier to recall items that should be measured for learning.

[0056] And as one of the important methods to achieve this is the process sheet in the above-described embodiment. Manufacturing engineers and the like can easily set various conditions and various variables for learning by simply performing simple operations while organizing the manufacturing process by making use of their own specialized knowledge and listing the management items in each process.

[0057] Also, the process sheet of the above-described embodiment may be useful in terms of simply visualizing the flow of the entire manufacturing process in an easy-to-understand manner. That is, in the conventional manufacturing industry, something like an easy-to-understand process sheet is not necessarily used, and mostly only the work procedures and the like are standardized and described as documents. The process sheet in the above-described embodiment is considered useful in that it makes it easier for engineers and the like to simply organize and manage the flow of the entire manufacturing process.

[0058] Also, although not described in the above embodiments, the criteria by which this system classifies the entire manufacturing process into each process may adopt any criteria. For example, this system may classify the entire manufacturing process into each process based on criteria such as "the state of the manufactured product changes", "data measurement is performed", "the manufactured product moves", "the work purpose changes", etc.

[0059] Also, although not described in the above embodiments, this system may be equipped with a function of automatically selecting the final inspection result value when the final inspection target value is selected by the user. Conversely, this system may be equipped with a function of automatically selecting the final inspection target value when the final inspection result value is selected by the user.

[0060] Also, although not described in the above embodiments, this system may be equipped with a function of changing the content of the learning process according to the relationship between the final inspection target value and the final inspection result value. Specifically, for example, when the final inspection target value and the final inspection result value are too far apart, the accuracy of learning may decrease in some cases. Therefore, this system may be equipped with a function of not using such data for learning when there is data in which the final inspection target value and the final inspection result value are too far apart.

[0061] Also, in the above embodiments, it was described that the user uploads the raw data, but it is not limited thereto. For example, the raw data may be uploaded by another third party or automatically acquired by this system.

[0062] Also, in the above embodiments, it was described that learning data is generated after performing predetermined preprocessing on the raw data, but it is not limited thereto. The preprocessing of the raw data is not necessarily performed and is optional.

[0063] Also, in the above-described embodiments, although the process table (especially the settings for each process) has been described as basically being arbitrarily created by the user, it is not limited thereto. For example, the system may be provided with a function of automatically generating a process table based on the content of the raw data uploaded by the user or the like.

[0064] Also, in the above-described embodiments, although the system has been described as being able to freely create a process table by the operation of the user, it is not limited thereto. The system may provide certain restrictions or rules in the process of having the user create a process table.

[0065] Also, in the above-described embodiments (especially the embodiments of FIGS. 6 and 7), although the system has been described such that the data items set in the process before the data item selected in FIG. 6 are set as the data necessary for learning, and the data items set in the process after the data item selected in FIG. 6 are set as unusable data items, it is not limited thereto. Such a specification of the system is merely an example, and the system may adopt any other specification.

[0066] Also, in the above-described embodiments, the allowable error has been described as the range (width) regarding whether the error between the measured value and the predicted value of the manufacturing condition variable can be regarded as the same manufacturing condition in consideration of the user's prior knowledge, subjectivity, etc. However, this definition may vary according to the specification of the system and the mode of the service to be applied. For example, the system may define the allowable error as the range (width) regarding whether the error between the measured value and the predicted value of the specification variable can be regarded as the same specification in consideration of the user's prior knowledge, subjectivity, etc.

[0067] Also, in the above-described embodiments (especially the embodiment of FIG. 6), although the user has been described as selecting the data item to be the target of interest from the data items included in "processing", it is not limited thereto. For example, the system may accept any data item included in "intermediate state" or "final state" as the target data item, or may accept, for example, a plurality of data items as the target data item.

[0068] Also, in the above-described embodiments (particularly the embodiment of FIG. 7), the content of the display for classifying data items (for example, "data necessary for learning", "data of the target to be known", etc.) is illustrative and not limited. For example, the present system may display any other criteria such as "cause", "cause (explanatory variable)", "explanatory variable", "result", "result (objective variable)", "objective variable", etc. as classification.

[0069] Also, although the explanation was omitted in the above-described embodiments, the present system may accept the selection of a plurality of data items, for example, when a plurality of data items are included in the process selected by the user.

[0070] Also, in the above-described embodiments (particularly the embodiment of FIG. 4), the present system has been described as inputting arbitrary values to the "final inspection target value" and the "final inspection result value", but it is not limited. For example, the present system may omit the input from the user for the "final inspection target value" and the "final inspection result value".

[0071] Also, in the above-described embodiments, the present system has been described as calculating the correlation coefficient between data items, but it is not limited. The present system may calculate any coefficient different from the correlation coefficient related to the correlation relationship, or the present system may calculate only any index not related to the correlation relationship.

[0072] Also, the present system may be provided with a function of generating a learning result with a reduced risk of overfitting by using a neural network or the like taking into account the content of the process table and the like.

[0073] Also, the present system may be provided with a function of outputting the completed process table in a form such as an image.

[0074] Also, the present system may be provided with a function of enhancing the completion degree of the process table by adding pseudo variables.

[0075] For example, the above-described series of processes can be executed by hardware or by software. That is, the functional configuration of FIG. 3 is merely an example and is not limiting. It suffices that the information processing system is provided with a function capable of executing the above-described series of processes as a whole, and there are no particular limitations on what functional blocks are used to realize this function. Also, the location of the functional blocks is not limited to the example of FIG. 3 and may be arbitrary. Furthermore, one functional block may be constituted by hardware alone, by software alone, or by a combination thereof.

[0076] When the series of processes is to be executed by software, the program constituting the software may be installed in a computer or the like from a network or a recording medium. The computer or the like may be a computer incorporated in dedicated hardware. Also, the computer or the like may be a computer capable of executing various functions by installing various programs, such as a general-purpose smartphone or personal computer in addition to a server.

[0077] A recording medium containing such a program may be constituted not only by a removable medium (not shown) distributed separately from the apparatus main body for providing the program to the user, but also by a recording medium or the like provided to the user in a state pre-installed in the apparatus main body.

[0078] Note that, in this specification, the steps of describing the program recorded on the recording medium include not only processes that are performed in time series in accordance with the order, but also processes that are executed in parallel or individually, even if they are not necessarily processed in time series. That is, some of the steps in the steps of FIG. 9 may be appropriately changed or omitted.

[0079] Moreover, the number and users of various hardware components constituting this system are arbitrary, and it may also include other hardware and the like in its configuration. Furthermore, for example, this system may cause some or all of the above-described series of processes to be executed by another server or the like managed in the cloud or the like.

[0080] In other words, it suffices that the information processing apparatus to which the present invention is applied has the following configuration, and various embodiments can be adopted. That is, the information processing apparatus to which the present invention is applied is an information processing apparatus that can be used for statistical learning processing related to the manufacture of articles, a process table generation unit (for example, process table generation unit 62) that generates a process table, which is an image related to the manufacturing process of the article, based on an input from a user; a data item setting unit (for example, data item reception unit 90) that sets data items related to the learning for each of the processes included in the process table; a learning condition determination unit (for example, learning condition determination unit 64) that determines the conditions for the learning process based on the contents of the set data items; a learning processing unit (for example, learning processing unit 68) that performs the learning process based on the learning conditions determined by the learning condition determination unit and outputs the result of the learning process; is sufficient.

[0081] Moreover, the learning condition determination unit may further include a non-use data item reception unit (for example, non-use data item reception unit 92) that receives a selection of data items not used in the learning process.

[0082] Moreover, this information processing apparatus may further include a correlation coefficient processing unit (for example, correlation coefficient processing unit 66) that calculates the correlation coefficient of each of the data items.

[0083] Further, the learning processing unit may output the result of the learning processing by preferentially using the data items subject to processing among the data items subject to processing and the data items subject to measurement.

Explanation of Signs

[0084] <Server> 1 ··· Server 11 ··· Control Unit 60 ··· Learning Data Management Unit 62 ··· Process Table Generation Unit 80 ··· Data Item Selection Reception Unit 82 ··· Process Table Presentation Unit 64 ··· Learning Condition Determination Unit 90 ··· Data Item Reception Unit 92 ··· Unused Data Item Reception Unit 66 ··· Correlation Coefficient Processing Unit 68 ··· Learning Processing Unit 300 ··· Learning Data DB 400 ··· Learning Result DB <User Device> 2 ··· User Device

Claims

1. An information processing apparatus that can be used for statistical learning processing related to the manufacture of an article, A process table presentation unit that presents a process table, which is an image related to the manufacturing process of the article and data items corresponding to each of the manufacturing processes, to the user; A process table changing means for changing the content of the process table according to the content of the first input operation when the first input operation by the user who has confirmed the process table is received; A process table determination means for determining the content of the process table based on the input operation of the user; A learning condition determination unit that determines the learning conditions based on the content of the data items corresponding to each of the process tables determined by the process table determination means and the content of the second input operation related to the setting of the learning processing conditions of the user; A learning processing unit that performs the learning processing based on the learning conditions determined by the learning condition determination unit and outputs the result of the learning processing; Comprising, The data items include data items arbitrarily set by the user, Information processing apparatus.

2. The learning condition determination unit further includes a non-use data item reception unit that receives the selection of data items not used in the learning process, The information processing apparatus according to claim 1.

3. Further comprising a correlation coefficient processing unit that calculates the correlation coefficient of each of the data items, The information processing apparatus according to claim 1.

4. The learning processing unit preferentially uses the data items related to processing among the data items related to processing and the data items related to measurement, and outputs the result of the learning processing, The information processing apparatus according to claim 1.

5. The learning condition determination unit can set any of the data items related to processing and the data items related to measurement as target variables, The information processing apparatus according to claim 1.

6. An information processing method executed by an information processing apparatus that can be used for statistical learning processing related to the manufacture of an article A process table presentation step of presenting a process table, which is an image related to the manufacturing process of the article and data items corresponding to each of the manufacturing processes, to the user; A process table change step of changing the content of the process table according to the content of the first input operation when the first input operation by the user who has confirmed the process table is received; A process table determination step of determining the content of the process table based on the input operation of the user; A learning condition determination step of determining the conditions of the learning process based on the content of the data items corresponding to each of the process sheets determined in the process sheet determination step and the content of the second input operation related to the setting of the conditions of the learning process of the user; A learning process step of performing the learning process based on the learning conditions determined in the learning condition determination step and outputting the result of the learning process; comprising; The data items include data items arbitrarily set by the user. An information processing method.

7. A program executed by an information processing apparatus that can be used for statistical learning processing related to the manufacture of an article, A process sheet presentation step of presenting to the user a process sheet that is an image related to the manufacturing process of the article and the data items corresponding to each of the manufacturing processes; A process sheet change step of changing the content of the process sheet according to the content of the first input operation when the first input operation by the user who has confirmed the process sheet is received; A process sheet determination step of determining the content of the process sheet based on the input operation of the user; A learning condition determination step of determining the conditions of the learning process based on the content of the data items corresponding to each of the process sheets determined in the process sheet determination step and the content of the second input operation related to the setting of the conditions of the learning process of the user; A learning process step of performing the learning process based on the learning conditions determined in the learning condition determination step and outputting the result of the learning process; causing to execute a process including; The data items include data items arbitrarily set by the user. A program.

Images